Ethical review and approval were not required for the study on human participants in accordance with local legislation and institutional requirements. Written informed consent from the participants’ legal guardian or next of kin was not required to participate in this study in accordance with national legislation and institutional requirements. The data used for the analysis had been deidentified and made public by the US FDA.

The FAERS is a repository of AE cases sent to the FDA through MedWatch (form 3500/3500a) [22,23]. The cases include AEs submitted voluntarily by health care professionals, individuals, and legal representatives and spontaneous mandatory reports by manufacturers and sponsors. At the time of the analysis, the FAERS contained 18,274,795 reports from January 2004 to September 2022.

The data sets have been deidentified and made available on the web [24].

The FAERS quarterly data sets were initially downloaded in text format. Due to the variability of data structure between quarters and the paucity of some of the variables, the cases were standardized to fit a uniform structure [25-27].

Out of a total of 18,274,795 reports, AE cases where the report was submitted by a health care professional (pharmacists, physicians, nurses, or other health care professionals) were selected as the initial data set (n=6,360,489). This allowed for minimizing reporting bias and adding to the clinical relevance of the studied cases. Further, reports where GLP-1 RAs were the only drug reported (further referred to as monotherapy) were selected as the GLP-1 RA cohort (n=17,653) to avoid potential confounding effects from concomitant medications. The cohort was further split into individual GLP-1 RA subcohorts: semaglutide (n=3230), dulaglutide (n=3768), exenatide (n=4493), and liraglutide (n=6162). SGLT-2 inhibitor monotherapy reports (canagliflozin, dapagliflozin, and empagliflozin; n=14,102) were selected as a control cohort. SGLT-2 inhibitors were selected as a control due to the comparable recommendation by diabetes management guidelines. Metformin monotherapy was initially selected as the control cohort to match the population of the cohort of interest (n=8536); however, only a single case of interest (Preferred Term [PT] code: thyroid cancer) was reported. The undetectable baseline of this AE made it impossible to use metformin as a control (Table 1). All the thyroid-related AE PT codes based on standard MedDRA queries and FDA Medical queries were used in the case selection process. Disproportionality analysis using reporting odds ratios (RORs) and 95% CIs was used to determine the statistical significance of the results. Of the 31,755 included reports, only those from health care professionals (pharmacists: n=3620, 11.4%; physicians: n=20,133, 63.4%; and other health care professionals: n=8002, 25.2%) were included in the analysis. The reports were primarily from the United States (United States: n=26,452, 83.3%; Japan: n=603, 1.9%; United Kingdom: n=540, 1.7%; France: n=476, 1.5%; Canada: n=349, 1.1%; Australia: n=349, 1.1%; and other counties: n=2986, 9.4%, <1% in each country). The reports were from the following years: 2010 (n=2128, 6.7%), 2011 (n=3430, 10.8%), 2012 (n=2096, 6.6%), 2013 (n=1143, 3.6%), 2014 (n=1080, 3.4%), 2015 (n=1969, 6.2%), 2016 (n=1429, 4.5%), 2017 (n=1651, 5.2%), 2018 (n=2159, 6.8%), 2019 (n=2191, 6.9%), 2020 (n=4096, 12.9%), 2021 (n=4954, 15.6%), and 2022 (n=3430, 10.8%; only the first 3 quarters at the time of the analysis).

In this study, we observed a reported association of GLP-1 RA treatment with thyroid hyperplasia and neoplasm AEs. To our knowledge, this is the first analysis of the FAERS to generate an ROR profile of GLP-1 RA monotherapy as a class and further analyze the individual GLP-1 RA monotherapy AEs compared to SGLT-2 inhibitors monotherapy AEs. The mean ages of the specific drug cohorts with these AEs were similar for comparison, ranging from 53.0 to 64.7 years. The thyroid hyperplasia and neoplasm AE association was significant for every GLP-1 RA in the class, with an even narrower range of 95% CI in the combined cohort. The RORs ranged from 7.65 to 35.92, with the lowest bound being 3.17. The highest mean ROR value was observed for liraglutide, and the lowest mean ROR was observed for exenatide, with only a small overlap in the 95% CIs. Interestingly, this trend was also seen in a similar analysis performed by Mali et al [20] using the European pharmacovigilance database (EudraVigilance), where this association was the strongest in the liraglutide cohort, followed by the exenatide cohort, with proportional reporting ranges of 27.5 (95% CI 22.7-33.3) and 22.5 (95% CI 17.9-28.3), respectively. The differences in the numbers and 95% CI ranges are due to the numbers of reports and type of analysis.

An association between T2DM and thyroid function has been previously established [28]. These disease states are often comorbid, and one affects the disease progression of the other [29]. However, the potential molecular mechanisms responsible for thyroid effects of GLP-1 RAs are insufficiently characterized and may include multiple pathways such as phosphoinositol-3 kinase and Akt serine/threonine kinase pathways; mitogen-activated protein kinase and extracellular signal-regulated kinase pathways; expression of GLP-1 receptors by C-cells; and GLP-1 association with triiodothyronine levels [21,30]. Thyroid hormone plays a pivotal role in nearly every aspect of lipid metabolism [31,32]. Thus, it was expected to observe thyroid hyperplasia and neoplasm–related AEs across all the T2DM drug cohorts that were investigated. However, only a single report of thyroid hyperplasia and neoplasm AE (thyroid cancer) was observed in the metformin monotherapy cohort, and only 7 were observed in the SGLT-2 inhibitor monotherapy cohort that was selected as the control due to similar T2DM disease-stage treatment guideline recommendations. In contrast, almost 200 of these AEs were reported for GLP-1 RAs in similarly sized cohorts, resulting in a statistically significant reported association. Therefore, as the GLP-1 RAs expand into non-T2DM indications such as obesity metabolic syndrome and other related conditions, controlled studies and better understanding of the molecular mechanisms of action are necessary to investigate this association and prevent potential serious consequences.

Since this is an association study, the causality between the thyroid hyperplasias and neoplasms and GLP-1 RAs was not clinically adjudicated. However, this analysis of a large population–scale AE database provides a strong signal that may be clinically significant. The numbers of AEs presented in the study do not represent all treated patients due to voluntary submissions resulting in over- and underreporting [33,34]. Additionally, due the nature and long progression of the studied AEs, often requiring invasive procedures for a proper diagnosis, the cases may be significantly underreported. The case narratives with additional information such as a thorough medical history and test or diagnostic results were not available in the data sets provided by the FDA due to privacy concerns. Other limitations include the possible over-the-counter medications and supplements that are often not reported to health care professionals and may potentially add noise to the cohort compositions, AE frequencies, and to a lesser extent, RORs and the respective CIs. However, considering the large number of individuals in the GLP-1 RA and control cohorts, which were matched by indication, the signal was statistically significant.